Tech lead for "wakeword cloud modeling", which mainly involves verifying the decision of an Alexa device to wake up, to prevent it from responding to false wakes and improve the customer experience.

Algorithm development on the wake word science team.

First year: Development of software for the speech recognition engine used in the Amazon Echo and other Amazon services.Second year: Algorithm development on the wake word science team.

Technical lead for a software system that processes optical sensor data to detect and track satellites, debris, and asteroids, currently in operational use. Advised leadership and sponsors as domain expert. Coordinated 4-7 person team to update software and extend it to new sensor systems. Designed, implemented, tuned, and tested algorithms and software on simulated and real data sets. Developed novel algorithms for fusion of satellite observations from diverse sensors.

Designed, implemented, tuned, and tested algorithms for processing optical sensor data to detect and track satellites, debris, and asteroids. Designed the software (in Java) for maximal reuse to optimize the sensitivity and accuracy of a diverse array of sensors. Supported deployment of software for operational use in a satellite ground system with rapid response analysis, problem diagnosis, and software patches. Designed and implemented novel algorithms for highly accurate discrimination of real object detections from sensor artifacts in real datasets. Presented algorithms and results to team, sponsors, and larger community in seminars and conferences.

Designed and implemented software in C++ for coordinating processing across a cluster for tracking and discrimination in missile defense applications. Designed and implemented algorithms in Java, Matlab, and IDL for statistical pattern recognition and pattern matching. Participated in live missile defense tests.